Study On Proportional-Integral And Model Predictive Control For Reactive Distillation And Reactive Dividing Wall Colum

Distillation is the most widely employed separation technology in chemical process,However,it is energy and capital intensive.Therefore,the energy requirement reduction in separation is of vital importance for chemical process.Process intensification has been widely investigated and applied in chemical industry for its advantages in energy consumption,equipment investment cost and equipment volume reduction.As the typical representative of process intensification technology,reactive distillation(RD)and reactive dividing wall column(RDWC)have received attention by both academic research and industrial application.RD and RDWC have advantages in reducing the number of equipment and decreasing the operation cost through integrating reaction and multiproduct separation process into one operating unit.However,due to the complexity of the integrated distillation process structure,the number of manipulated variables increases.Besides,the interaction between controlled and manipulated variables is highly nonlinear,which requires process control should be effectively addressed.In the present work,conventional RD process and RDWC of formic acid production were investigated in detail.Firstly,Aspen Plus software was used to simulate the conventional RD process and RDWC process with heterogeneous kinetics.Then genetic algorithm(GA)was coupled with rigorous simulations to optimize both integer and continuous parameters of conventional RD process and RDWC with total annualized cost(TAC)as objective,which shows to be effective and robust for finding the optimal solution.Secondly,multi-loop temperature PI control structure was developed and advanced controller-model predictive control scheme was designed,respectively for conventional RD process and RDWC.The weights of MPC controller were optimized by multi-objective genetic algorithm.Finally,dynamic responses of temperature PI control and model predictive control for conventional RD process and RDWC were obtained and compared subject to various feed disturbances.By comparing dynamic performance of the two control structures,it is found that both PI control and model predictive control could return to steady state for the series of feed disturbance.However,the maximum deviation in dynamic response under PI control reaches an unacceptable level for some process variables,which could be significantly decreased by effective MPC.In terms of a quantitative index such as maximum deviation,ISE etc.,the dynamic response could be significantly improved by employing MPC,which confirmed the superiority of MPC in controlling this highly coupled system.